neurosurgical focus Neurosurg Focus 39 (3):E9, 2015

4.7-T diffusion tensor imaging of acute traumatic peripheral nerve

Richard B. Boyer, PhD,1,2 Nathaniel D. Kelm, MS,1,3 D. Colton Riley, BS,2 Kevin W. Sexton, MD,2,4 Alonda C. Pollins, MLI,2 R. Bruce Shack, MD,2 Richard D. Dortch, PhD,1,3 Lillian B. Nanney, PhD,2 Mark D. Does, PhD,1,3 and Wesley P. Thayer, MD, PhD1,2

Departments of 1Biomedical Engineering and 2Plastic Surgery, Vanderbilt University Medical Center; 3Vanderbilt University Institute of Imaging Science, Nashville, Tennessee; and 4Department of Surgery, University of Arkansas for Medical Sciences, Little Rock, Arkansas

Diagnosis and management of peripheral is complicated by the inability to assess microstructural features of injured nerve fibers via clinical examination and electrophysiology. Diffusion tensor imaging (DTI) has been shown to accurately detect nerve injury and regeneration in crush models of peripheral nerve injury, but no prior studies have been conducted on nerve transection, a surgical emergency that can lead to permanent weakness or paralysis. Acute sciatic nerve were performed microsurgically to produce multiple grades of nerve transection in rats that were harvested 1 hour after surgery. High-resolution diffusion tensor images from ex vivo sciatic nerves were obtained using diffusion-weighted spin-echo acquisitions at 4.7 T. Fractional anisotropy was significantly reduced at the injury sites of transected rats compared with sham rats. Additionally, minor eigenvalues and radial diffusivity were profoundly elevated at all injury sites and were negatively correlated to the degree of injury. Diffusion tensor tractography showed discontinui- ties at all injury sites and significantly reduced continuous tract counts. These findings demonstrate that high-resolution DTI is a promising tool for acute diagnosis and grading of traumatic peripheral nerve injuries. http://thejns.org/doi/abs/10.3171/2015.6.FOCUS1590 Key Words diffusion tensor imaging; diffusion tensor tractography; MRI; neurography; peripheral nerve injury; nerve transection; neurotmesis

eurotmesis, or peripheral nerve transection, is a fourth- or fifth-degree peripheral nerve lesion is critical to common but difficult to distinguish diagnosis fol- recovery.9 Failure to microsurgically repair a nerve in a lowing traumatic injury. Electrodiagnostic studies timely manner may result in irreversible muscle atrophy, andN clinical examination are standard practice for peripher- weakness, paralysis, or formation of painful traumatic neu- al nerve assessment, but neither can perfectly discriminate romas.9 a severe nerve laceration from a self-resolving axonotmetic Peripheral nerve injury is one of the few acute neu- or neurapraxic injury in the acute setting.16 Additionally, rological disorders in which imaging is not the standard may take up to 4 weeks before of care. Transected nerves can be identified accurately these injuries can be accurately diagnosed in the absence of with the aid of high-resolution ultrasound.10,29 However, exploratory surgery.14 This is also true for peripheral nerve traumatic injuries have proven difficult to diagnose using repair, in which it may take up to a year, depending on the ultrasonography, because of large hematomas, extensive site and severity of injury, before targets are reinnervated skin lacerations, , and disruption of the normal anat- and surgical outcome is known. Although a “watch and omy.10 Furthermore, no studies have shown the ability of wait” approach might appear to be the optimal conserva- high-resolution ultrasonography to reveal early nerve re- tive management, the time following a severe Sunderland generation, which ultimately would indicate whether addi-

Abbreviations AUC = area under the curve; DTI = diffusion tensor imaging; DTPA = diethylene triamine pentaacetic acid; FA = fractional anisotropy; MD = mean diffu- sivity; MRN = MR neurography; PBS = phosphate-buffered saline; PFA = paraformaldehyde; ROC = receiver-operator characteristic; ROI = region of interest; SNR = signal- to-noise ratio; l|| = axial diffusivity; l^ = radial diffusivity. submitted February 26, 2015. accepted June 16, 2015. include when citing DOI: 10.3171/2015.6.FOCUS1590.

©AANS, 2015 Neurosurg Focus Volume 39 • September 2015 1

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC R. B. Boyer et al. tional surgical intervention is required.28 MR neurography tinguish acute peripheral nerve transection lesions from (MRN) leverages the microstructural features of axonal intact nerve in the absence of macroscopic discontinuity. bundles to produce enhancement of peripheral nerves in We tested this hypothesis using a complete and partial rat MRI.6,12,15,17,26 Nerves normally exhibit a moderate signal sciatic nerve injury model and high-resolution DTI at 4.7 on T2-weighted sequences, but they become hyperintense T. The partial nerve transection injury model was chosen following injury, because of prolongation of the T2 relax- to determine the correlation of axonal disruption and DTI ation time.1 Although this can be a sensitive marker of pe- parameters in the acute injury setting. This model was ripheral nerve pathology, conventional T2-weighted MRI also chosen to mimic experimental axonal-level nerve re- lacks specificity for type and severity of nerve injury.5,7 pair, such as polyethylene glycol fusion, in which a subset In the most severe cases of nerve injury involving gaps, of proximal and distal stumps are believed to recover MRN is capable of detecting nerve discontinuity.11 How- anatomical and functional connection immediately fol- ever, more than half of all high-grade nerve transections lowing treatment.18,27 Immediately following injury, sciat- have minimal or no gap present. ic nerves were microsurgically repaired to maintain close Diffusion tensor imaging (DTI) is an MRI sequence approximation of the epineurium, mimicking a lesion that that leverages the anisotropic diffusion of water molecules would be difficult to detect with conventional MRI. In ad- through to aid visualization of neural tracts and dition to determining the diagnostic capabilities of DTI in assessment of microstructural features. DTI parameters, high-grade nerve lesions, this study provides a baseline for such as fractional anisotropy (FA), mean diffusivity (MD), postoperative evaluation of peripheral nerve repairs using and individual eigenvalues (l1–l3) are able to quantify DTI. microstructural characteristics of nerves, such as axon density and thickness.35 In the CNS, DTI has been Methods particularly useful in detecting axonal pathology in trau- Experimental Design matic injury and .2,8 Multiple inves- tigators have evaluated DTI for detecting peripheral nerve The objective of this study was to evaluate the accu- crush and traction nerve injuries, which show significant racy of DTI parameters, including fractional anisotropy changes in FA following injury.22,23,25,31 However, limited (FA), mean diffusivity (MD), radial diffusivity (l^), and data exist on high-grade nerve lacerations and there are individual eigenvalues in detecting acute peripheral nerve no prior animal studies of DTI following microsurgical transection. This study was performed using a rat sciatic coaptation. nerve injury model with sham (n = 6), partial (n = 9 [n = 3 The ability to noninvasively assess axonal integrity per group]), and complete transection injuries (n = 6). DTI could have vast implications in the management of pe- was performed on preparations of excised sciatic nerve ripheral nerve disorders. In the case of traumatic nerve and amputated hind limbs following fixation. injury, an accurate noninvasive measurement of axonal continuity could assist in surgical planning and postop- Animal Model erative management.15 Iatrogenic peripheral nerve injury Adult female Sprague-Dawley rats aged 10–12 weeks may also occur during certain medical procedures, such (200–250 g; Charles River Lab) were used in the experi- as nerve blocks, leading to an expensive and time-con- ments. Rats were housed in a central animal care facility suming neurological and musculoskeletal workup.19 DTI and provided food and water ad libitum. All study pro- may be able to quickly diagnose these conditions and ex- cedures involving animals were approved by our institu- pedite therapy. Last, there has been growing interest in tional animal care and use committee. neuromodulatory therapy, which leverages the peripheral nerve tracts as targets for electrophysiological interven- Sciatic Nerve Injury tions. As peripheral nerve interventions are developed, it In each rat, the left hindquarter was shaved and the lat- will be important to have noninvasive tools for diagnosis eral thigh was incised from the sciatic notch to the popli- and management of adverse events (e.g., needle injuries). teal fossa. The biceps femoris was reflected to expose the The specificity of DTI for axonal tracts makes it an ideal sciatic nerve. Surrounding fascia was sharply dissected tool for guiding these cases. and the sciatic nerve mobilized. The injury was created by Peripheral nerve research may also benefit from DTI. completely or partially transecting the sciatic nerve with Nerve regeneration models are highly dependent on be- extra fine Vannas microsurgical scissors (RS-5640; Roboz havioral testing and histology, which require large animal Surgical Instrument Co.). In partial transections, approxi- trials with multiple end points and error-prone assess- mately 25% or 50% of the nerve diameter was advanced ments, such as the sciatic functional index.3,20 High-res- onto the cutting edge of the scissor under high magnifi- olution DTI may not only provide real-time, noninvasive cation. Sham nerves were mobilized and not transected. measurements of axon viability but also allow for several Microsurgical suturing of the epineurium was performed intermediate measurements of nerve recovery in survival on injured nerves to control for variable nerve gaps and studies. This could reduce the large number of animals to stabilize injury sites. The muscle was repaired and skin required in nerve regeneration studies, in addition to pro- incision closed in a subgroup of rats for in situ imaging viding greater understanding of the regeneration time line. (n = 8). All rats were killed at 1 hour after injury with As DTI can be used to accurately detect acute peripher- perfusion fixation using 4% paraformaldehyde (PFA) in al nerve crush and traction injuries, we hypothesized that phosphate-buffered saline (PBS). The sciatic nerve was high-resolution DTI also could be used to accurately dis- excised, straightened on a wood dowel, and incubated in

2 Neurosurg Focus Volume 39 • September 2015

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC Diffusion tensor imaging of acute nerve injury

4% glutaraldehyde/0.5% PFA in PBS at 4°C. Hindquarter applied to the nerve ending, using a 27-gauge needle. The amputation was performed on the subgroup of rats with dyed nerve ending was embedded in petroleum jelly to closed wounds and the amputated leg was incubated in 4% prevent leakage of the tracer and placed in PBS-sodium glutaraldehyde/0.5% PFA in PBS at 4°C. azide (0.02%). The nerve was stored at 37°C in the dark for 2 weeks to allow for tracing in the proximodistal direc- Tissue Sample Preparation tion through the injury site. After 24 hours of postfixation, excised nerves were Segments of proximal and distal nerve were postfixed placed in PBS plus 2 mM Gd-diethylene triamine pen- in 10% neutral buffered formalin for 24 hours and paraf- taacetic acid (DTPA) (Magnevist, Bayer HealthCare) at fin embedded. Formalin-fixed, paraffin-embedded tissues 4°C for at least 24 hours before imaging. For imaging, were sectioned at 5 mm, placed on slides, and warmed excised nerves were trimmed to approximately 1 cm in overnight at 60°C. Slides were deparaffinized with xylene length (with the injury site centered) and placed in glass and rehydrated with graded alcohols, ending in distilled capillary tubes (2-mm outer diameter) filled with a per- water. Prolong Gold antifade reagent with DAPI (Life fluropolyether liquid (Fomblin, Solvay Solexis) for sus- Technologies/Thermo Fisher Scientific) was applied to the ceptibility matching, preventing tissue dehydration, and a slide and a coverslip was placed over the slide, which was signal-free background. For higher throughput, 6 excised allowed to dry and was stored at 4°C until ready to view. sciatic nerves in a hexagonal arrangement were imaged Digital images of slides were acquired with a Nikon simultaneously. Hindlimb samples were postfixed for 1 AZ100 M widefield fluorescence microscope (Nikon week, followed by at least 1 week of washing in PBS plus Corp.). CM-DiI fluorescence was observed using a Cy-3 2 mM Gd-DTPA before imaging. For imaging, hind limbs fluorescence filter with 548-nm/561-nm excitation/emis- were placed in MR-compatible tubes filled with perfluo- sion wavelengths. All sections were viewed with a 5× ropolyether liquid. Zeiss Plan fluorescence objective (Carl Zeiss AG) without coverslips. Diffusion Tensor MRI Statistical Analysis MRI was performed on a 4.7-T, 31-cm horizontal bore, Agilent DirectDrive scanner (Agilent Technologies) using Prism 6 (GraphPad Software) was used for plots and a 38-mm Litz quadrature coil (Doty Scientific) for radio- statistical calculations. Quantitative region of interest frequency transmission and reception. For excised nerve (ROI) statistics were performed with Student t-tests cor- imaging, the FOV was 9.6 × 9.6 × 12 mm3 and the matrix rected for multiple comparisons using the Holm-Sidak size was 96 × 96 × 32 for a nominal resolution of 100 × method, with significance determined as at an a of 0.05. 100 × 375 mm3 (375 mm along the nerve). For hindlimb imaging, the FOV was 48.0 × 25.6 × 28.8 mm3 and the ma- Results trix size was 192 × 128 × 144, for a nominal resolution of Injury Validation 250 × 200 × 200 mm3 (250 mm along the nerve). DTI data were acquired using a 3D diffusion-weighted spin-echo Various degrees of high-grade peripheral nerve injury sequence with a TR/TE of 170/23.0 msec and 12 signal av- were evaluated in this study using complete and partial erages for excised nerves, and a TR/TE of 170/22.1 msec transections of rat sciatic nerves followed by ex vivo DTI. and 2 signal averages for hind limbs. Diffusion weighting The severity of nerve injury was predicted at time of was achieved with a d/D of 4/12 msec, a prescribed b value surgery with careful microsurgical nerve dissection and of 2000 sec/mm2, and 6 directions. One b = 0 image was verified with histology. CM-DiI membrane label was ob- acquired, for a total of 7 images in a scan time of approxi- served in all sham and transected nerves and, therefore, mately 12 hours. could not be used as a direct marker of injury severity, most likely due to tracer leakage or membrane approxi- DTI Postprocessing and Analysis mation. However, during evaluation of the proximal nerve sections, we observed that injured nerves had greater axon Image data reconstruction and diffusion tensor calcu- caliber compared with sham nerves. Completely transect- lations were performed using in-house written code in ed nerves were found to have significantly larger mean MATLAB (Mathworks). 3D image volumes were zero axon caliber compared with sham nerves (p < 0.01) (Fig. padded twice in each direction during reconstruction from 1C). As observed in diffuse axonal injury in brain white k-space data. Diffusion tensors were estimated voxelwise matter, cytotoxic edema occurs rapidly after damage to using a linear least-squares approach. From the diffusion the axolemmal membrane, causing axonal swelling.24 In tensor, DTI metrics, including FA and MD, axial diffusiv- this study, we observed that mean axon caliber also scaled ity (l||), and l^, were computed voxelwise. DTI tractogra- 21 with the predicted severity of nerve injury (Fig. 1D), with phy was performed using the ExploreDTI toolbox. moderate correlation of predicted percentage of injured axons and mean axon caliber (r2 = 0.59; p = 0.002). There Histology was wide variability in axon caliber in nerves predicted After completion of imaging, the distal end of the nerve to be 75% transected, which we suspect was due to surgi- was trimmed to provide a fresh stump and chlorometh- cal error. Removing the nerves that were 75% transected ylbenzamido-DiI membrane-labeling paste (NeuroTrace from the statistical analysis improved the correlation of CM-DiI Tissue Labeling Paste; Life Technologies/Ther- predicted injury severity and mean axon caliber (r2 = 0.82; mo Fisher Scientific), a carbocyanine dye derivative, was p = 0.0001).

Neurosurg Focus Volume 39 • September 2015 3

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC R. B. Boyer et al.

curve [AUC] 1.000; p < 0.001), but a larger sample size would be required to prevent Type II error. Table 1 shows the calculated sensitivities and specificities of different FA thresholds for diagnosing complete nerve transection. Continuous fiber tractography was performed with a maximum 45° angle and threshold FA of 0.49, correspond- ing to 1 SD above the mean FA of all injury sites. Continu- ous tracts were observed throughout the body of each sham nerve (Fig. 3A), whereas there were multiple discontinui- ties present in transected nerves (Fig. 3B), particularly at the injury site. Fewer tracts were visible in the proximal segments of injured nerves, most likely due to reduced FA from tissue edema. Additionally, the number of continuous tracts seeded proximally and passing through the injury site and distal ROI were significantly lower in transected nerves compared with sham rats (p < 0.01; Fig. 3C). Prin- cipal diffusion vectors at injury sites were also observed Fig. 1. Predicted nerve injury severity correlated with proximal axon to be more heterogeneous than sham nerves, with greater caliber. A and B: Representative fluorescent microscopy images of variability in diffusion direction (Fig. 4). CM-DiI membrane label in (A) sham and (B) completely transected sci- atic nerves. is visible as holes in the center of red membrane DTI of Partially Transected Nerves rings. C: Axon caliber was measured 3 mm proximal to the injury site for all axons visible in the axial cross-sections of sham and completely For determining the degree to which partial severance transected nerves (mean ± SEM; n = 6). D: Mean axon caliber cor- of the sciatic nerve could be distinguished with DTI, we related moderately with the predicted injury severity (r2 = 0.59; n = 15; imaged partially transected nerves that were graded at the p = 0.002). However, correlation strongly improved after removing the 2 time of surgery as 25%, 50%, or 75% nerve severance and nerves predicted to be 75% transected (r = 0.82; n = 12; p = 0.0001). validated histologically, as described under Methods. DTI Dashed lines indicate the 95% CI. **p < 0.01, unpaired t-test with Welch correction. parameters were evaluated at the injury ROI for each in- jured nerve and at the center of each sham nerve. Mean DTI of Completely Transected Nerves FA was significantly reduced in all injured nerves com- pared with sham nerves (Fig. 5A). However, there were no Diffusion tensor images of excised sciatic nerves were significant differences between the levels of injury. Mean evaluated at 3 ROIs: the injury site and 3 mm proximal FA of sham nerves was strongly correlated with the nerve and 3 mm distal to the injury site. To prevent partial vol- axon count, but this relationship was not present in injured ume effects, a margin of 100 mm was excluded from the nerves (Fig. 5B). Interestingly, no significant differences ROI. This also prevented the inclusion of suture material were found in MD (Fig. 5C) or l|| (Fig. 5E), as has been in ROI calculations, since knots are limited to the outer- observed in prior studies of predegenerated contusive most epineurium. FA, MD, and eigenvalues were mea- nerve injuries.31 Following Wallerian degeneration, sciatic sured for each ROI (Fig. 2A–C). As expected, the FA was nerves have been shown to have normal MD but mark- significantly decreased at the injury sites of transected edly decreased l|| and increased radial diffusivity.4 In this compared with sham rats (p < 0.01). Although not as study, minor eigenvalues (l2, l3) and radial diffusivity strong as at the injury site, the proximal ROI of transected were profoundly increased at all injury sites immediately nerves also had lower FA compared with sham rats (p < following injury (Fig. 5E). 0.01). Whereas the drop in FA at the injury site indicates Although there was no significant correlation between axonal loss, the proximal ROI changes were most likely eigenvalues and predicted degree of injury, there was a due to edema following nerve injury, causing an increase weak negative correlation between l2 and l^ to proximal in extracellular fluid compartment and an increase in axo- axon caliber (Fig. 5G). Neither parameter was correlated nal edema, which would be expected to decrease FA while simultaneously increasing MD and eigenvalues. Although to nerve axon count in sham or injured nerves. MD was significantly lower at the injury site (p < 0.01) and unchanged in the proximal and distal ROIs, the mi- DTI of Hind Limbs nor eigenvalues showed an increasing trend at all ROIs As proof of concept, ex vivo imaging was performed in transected nerves compared with sham rats (Fig. 2C). on amputated rat hind limbs with sham or complete tran- Additionally, radial diffusivity, calculated as (l2 + l3)/2, section injuries. Tractography was performed with mul- was significantly increased in the injury site (p < 0.05) and tiple seed points spaced 1 mm apart along the axial length proximal ROI (p < 0.05) of transected nerves compared of the sciatic nerve. We found that seeding from a single with sham rats. We observed larger variance in distal ROI axial slice, as was done in all excised nerves, resulted in quantitative measurements, which may be due to variation short tracts, even at low FA thresholds. By seeding from in fascicular branching at this level in the sciatic nerve. multiple axial slices along the nerve, we were able to cap- Receiver-operator characteristic (ROC) curves were eval- ture the length of the nerve while continuing to observe uated for injury-site DTI measurements (Fig. 2D). Injury- discontinuities at all injury sites (Fig. 6A and B). Quantita- site FA produces a near-perfect ROC curve (area under the tive DTI measurements were calculated at multiple ROIs

4 Neurosurg Focus Volume 39 • September 2015

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC Diffusion tensor imaging of acute nerve injury

Fig. 2. FA enabled detection of complete nerve transection with high sensitivity and specificity. A: FA was reduced at and proximal to injury sites (mean ± SEM; n = 6). B: MD was reduced only within injury sites (mean ± SEM; n = 6). C: Independent eigenvalues (l1, l2, l3) and l^ were measured at proximal, injury, and distal regions. Axial diffusivity (l||) was reduced within injury sites and l^ was increased at and proximal to injury sites (mean ± SEM; n = 6). D: ROC curve analysis of FA shows high sensitivity and specificity for detection of any nerve injury (AUC 1.00; n = 21; p < 0.001). Unpaired t-tests corrected for multiple comparisons using the Holm-Sidak method. *p < 0.05; **p < 0.01.

(i.e., the injury site and proximal [1, 3, and 5 mm] and asymptote at 3 mm, but remains significantly lower than distal [1, 3, and 5 mm] from the injury site). The additional in sham animals (p < 0.05). proximal and distal ROIs were included in this analysis to compare effects of surrounding tissue inflammation and edema. Similar to excised nerves, FA was significantly Discussion reduced at the injury site of transected nerves (p < 0.01) We present findings that show high-resolution DTI and, to a lesser extent, in the proximal and distal ROIs enables detection of acute peripheral-nerve severance in- (Fig. 6C). We hypothesized that external tissue effects juries in the absence of macroscopic nerve discontinuity. should be similar within the entire surgical site, whereas We also provide a baseline measurement of DTI param- sciatic nerve injury effects should diminish with increas- eters following microsurgical neurorrhaphy, which builds ing distance from the injury site. This was confirmed by a foundation for the use of DTI in postoperative manage- our measurements of FA at multiple distances from each ment. Importantly, intralesional FA was a highly sensitive injury site (Fig. 6D), in which the FA appears to reach an and specific marker of nerve injury even with partial nerve

TABLE 1. Sensitivity and specificity of FA for all nerve injuries FA Threshold Sensitivity, % 95% CI Specificity, % 95% CI Likelihood Ratio <0.5630 100.0 78.20–100.0 100.0 54.07–100.0 <0.6197 100.0 78.20–100.0 83.33 35.88–99.58 6.000 <0.6406 100.0 78.20–100.0 66.67 22.28–95.67 3.000 <0.6838 100.0 78.20–100.0 50.00 11.81–88.19 2.000 <0.7127 100.0 78.20–100.0 33.33 4.327–77.72 1.500 <0.7219 100.0 78.20–100.0 16.67 0.4211–64.12 1.200

Neurosurg Focus Volume 39 • September 2015 5

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC R. B. Boyer et al.

Fig. 3. Tract continuity diminished in completely transected nerves. A and B: Diffusion tensor tractography in sham (A) and completely transected (B) sciatic nerves. Blue lines indicate seed points for tractography located at 3 mm proximal and distal to the injury sites. C: Continuous tract counts through the injury region were significantly reduced in completely transected nerves (mean ± SEM; n = 6). Unpaired t-test with Welch correction. *p < 0.01. lacerations, and, unlike nerve crush injuries, there was a may be capable of both diagnosis and grading of traumatic concurrent increase in radial diffusivity. This could be peripheral nerve injuries. the result of immediate membrane and myelin disruption Prior studies evaluating DTI of peripheral nerve in- present in severance but not contusive nerve injuries. juries have focused on the effects of crush or contusion Radial diffusivity following injury was also correlated with blunt force trauma.23,25,31 Although nerve crush is a to the degree of nerve severance. If proximal axon cali- significant cause of morbidity, typically occurring as a sur- ber is an accurate reflection of axonal swelling following gical complication, traumatic nerve injuries often involve nerve injury, then this could indicate that l2 and l^ are partial or complete neurotmesis that can result in lifelong inversely proportional to the degree of nerve severance disability if not promptly diagnosed and microsurgically following injury. Although this is counterintuitive, since repaired.9 Providing a new method of detecting nerve tran- both parameters were significantly increased following section could prevent delays in surgical repair that lead to transection, it is likely that there are multiple microstruc- irreversible weakness, loss of function, and other undesir- tural factors contributing to these changes. One possible able sequelae. Additionally, DTI could be highly valuable explanation is that severance largely increases l2 and l^ as a tool for assessing nerve repairs and determining when because of membrane and myelin disruption while simul- a revision surgery is necessary. taneously, and to a lesser extent, decreasing because of im- There are multiple limitations in this study. Most im- mediate axonal swelling. These findings are supported by portantly, the majority of this study was conducted on ex- prior studies of DTI following sciatic nerve transection or cised sciatic nerves that had been cleaned and cross-linked traction injuries22,23 and suggest that high-resolution DTI in fixative prior to imaging. Fixation is known to reduce

Fig. 4. Principal diffusion vectors are heterogeneous at transection sites. Principal diffusion vector glyphs overlaying the axial FA map of the injury region in sham (upper) and completely transected (lower) nerves. Colors indicate vector direction. Lighter regions of the map indicate higher FA.

6 Neurosurg Focus Volume 39 • September 2015

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC Diffusion tensor imaging of acute nerve injury

Fig. 5. FA and l^ detected all partial injuries and l^ correlated with injury severity. Quantitative analysis of DTI parameters in sham, partial, and complete sciatic nerve transection injuries. A: FA was similarly reduced in the injury region of all complete and partial nerve injuries (mean ± SEM; n = 21). C: MD showed a nonsignificant increase in all injury groups. E: Minor eigenvalues and l^ were significantly increased in the injury region in all injury groups (mean ± SEM; n = 21). B, D, F, and H: There was no correlation of FA (B), MD (D), l|| (F), or l^ (H) measured at injury sites with the number of axons in injured nerves. G: l^ was negatively correlated to axon caliber in injured nerves and indirectly to injury severity (r2 = 0.357). One-way ANOVA corrected for multiple comparisons with Tukey test. *p < 0.05; #p < 0.01; $p < 0.001; ***p < 0.0001.

Neurosurg Focus Volume 39 • September 2015 7

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC R. B. Boyer et al.

Fig. 6. FA remained sensitive to nerve transection in ex vivo hind limbs. A and B: Ex vivo diffusion tensor tractography of rat hind limbs with sham (A) and completely transected (B) sciatic nerves. C: FA measured at injury and each region proximal and distal to the injury (mean ± SEM; n = 8). D: FA approached sham values with increasing distance from the injury site but remained significantly lower at all points (mean ± SEM; n = 8). Unpaired t-tests corrected for multiple comparisons using the Holm-Sidak method. *p < 0.05; **p < 0.01. isotropic diffusion, but should have minimal effect on injuries in a rat model. We have also provided an imme- relative tissue anisotropy.30 As observed in our hindlimb diate baseline of DTI parameters following microsurgical imaging, surrounding tissue and edema confounds the in coaptation of the sciatic nerve. Additional survival studies vivo nerve DTI signal. Although FA remained an accurate of traumatic peripheral nerve injury are needed to evalu- marker of traumatic nerve injury, tractography capabilities ate DTI as a tool for postoperative monitoring of nerve were limited and required a different approach for tract regeneration. seeding. Use of a multicompartment diffusion-MRI model such as diffusion basis spectrum imaging could isolate the Acknowledgments confounding effects of edema and inflammation and im- 32,33 We thank Dr. Alexander Leemans for providing ExploreDTI prove nerve assessment. software that was used for nerve tractography. This work was sup- A second limitation of this study was the time required ported by Public Health Service award T32 GM07347 from the for high-resolution image acquisition—up to 12 hours for National Institute of General Medical Studies for the Vanderbilt 1 set of 6 nerves. For translating this technology, it will Medical-Scientist Training Program, NIH/NIBIB grant EB001744, be critical to find the resolution and FOV that minimize and NIH training grant T32 EB014841. This work was also sup- scan time while maintaining signal-to-noise ratio (SNR) ported by Department of Defense grant OR120216. and diagnostic accuracy. New high-resolution DTI meth- ods such as outer-volume suppression and elliptically refo- References cused zonally oblique multislice have been developed that 1. Aagaard BD, Lazar DA, Lankerovich L, Andrus K, Hayes leverage reduced-FOV imaging to significantly decrease CE, Maravilla K, et al: High-resolution magnetic resonance 13,34 data-set sizes and scan times up to 8-fold. Sensitivity imaging is a noninvasive method of observing injury and encoding and other parallel imaging techniques can be recovery in the peripheral nervous system. Neurosurgery used in combination with reduced-FOV techniques to fur- 53:199–204, 2003 ther reduce data-set sizes, but their use is limited by SNR 2. Agosta F, Absinta M, Sormani MP, Ghezzi A, Bertolotto reduction. A, Montanari E, et al: In vivo assessment of cervical cord damage in MS patients: a longitudinal diffusion tensor MRI study. Brain 130:2211–2219, 2007 Conclusions 3. Bain JR, Mackinnon SE, Hunter DA: Functional evaluation of complete sciatic, peroneal, and posterior tibial nerve le- The benefits of DTI for noninvasive assessment of sions in the rat. Plast Reconstr Surg 83:129–138, 1989 axonal pathology have been well established in the cen- 4. Beaulieu C, Does MD, Snyder RE, Allen PS: Changes in tral and peripheral nervous systems. This study provides water diffusion due to Wallerian degeneration in peripheral confirmation that high-resolution DTI is capable of aiding nerve. Magn Reson Med 36:627–631, 1996 acute diagnosis and grading of traumatic peripheral nerve 5. Behr B, Schnabel R, Mirastschijski U, Ibrahim B, Angenstein

8 Neurosurg Focus Volume 39 • September 2015

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC Diffusion tensor imaging of acute nerve injury

F, Schneider W: Magnetic resonance imaging monitoring of tion by diffusion tensor imaging. J Magn Reson Imaging peripheral nerve regeneration following neurotmesis at 4.7 33:535–542, 2011 Tesla. Plast Reconstr Surg 123:1778–1788, 2009 26. Pham M, Wessig C, Brinkhoff J, Reiners K, Stoll G, Bend- 6. Bendszus M, Stoll G: Technology insight: visualizing pe- szus M: MR neurography of sciatic nerve injection injury. J ripheral nerve injury using MRI. Nat Clin Pract Neurol Neurol 258:1120–1125, 2011 1:45–53, 2005 27. Riley DC, Bittner GD, Mikesh M, Cardwell NL, Pollins AC, 7. Bendszus M, Wessig C, Solymosi L, Reiners K, Koltzenburg Ghergherehchi CL, et al: Polyethylene glycol-fused allografts M: MRI of peripheral nerve degeneration and regeneration: produce rapid behavioral recovery after ablation of sciatic correlation with electrophysiology and histology. Exp Neurol nerve segments. J Neurosci Res 93:572–583, 2015 188:171–177, 2004 28. Simon NG, Kliot M: Diffusion weighted MRI and tractogra- 8. Bennett RE, Mac Donald CL, Brody DL: Diffusion tensor phy for evaluating peripheral nerve degeneration and regen- imaging detects axonal injury in a mouse model of repetitive eration. Neural Regen Res 9:2122–2124, 2014 closed-skull . Neurosci Lett 513:160– 29. Suk JI, Walker FO, Cartwright MS: Ultrasonography of pe- 165, 2012 ripheral nerves. Curr Neurol Neurosci Rep 13:328, 2013 9. Campbell WW: Evaluation and management of peripheral 30. Sun SW, Neil JJ, Liang HF, He YY, Schmidt RE, Hsu CY, et nerve injury. Clin Neurophysiol 119:1951–1965, 2008 al: Formalin fixation alters water diffusion coefficient magni- 10. Cartwright MS, Chloros GD, Walker FO, Wiesler ER, tude but not anisotropy in infarcted brain. Magn Reson Med Campbell WW: Diagnostic ultrasound for nerve transection. 53:1447–1451, 2005 Muscle Nerve 35:796–799, 2007 31. Takagi T, Nakamura M, Yamada M, Hikishima K, Momoshi- 11. Chhabra A, Andreisek G, Soldatos T, Wang KC, Flammang ma S, Fujiyoshi K, et al: Visualization of peripheral nerve AJ, Belzberg AJ, et al: MR neurography: past, present, and degeneration and regeneration: monitoring with diffusion future. AJR Am J Roentgenol 197:583–591, 2011 tensor tractography. Neuroimage 44:884–892, 2009 12. Filler AG, Howe FA, Hayes CE, Kliot M, Winn HR, Bell BA, 32. Wang X, Cusick MF, Wang Y, Sun P, Libbey JE, Trinkaus K, et al: Magnetic resonance neurography. Lancet 341:659–661, et al: Diffusion basis spectrum imaging detects and distin- 1993 guishes coexisting subclinical inflammation, demyelination 13. Gaggl W, Jesmanowicz A, Prost RW: High-resolution re- and axonal injury in experimental autoimmune encephalomy- duced field of view diffusion tensor imaging using spatially elitis mice. NMR Biomed 27:843–852, 2014 selective RF pulses. Magn Reson Med 72:1668–1679, 2014 33. Wang Y, Wang Q, Haldar JP, Yeh F-C, Xie M, Sun P, et al: 14. Gaudet AD, Popovich PG, Ramer MS: Wallerian degenera- Quantification of increased cellularity during inflammatory tion: gaining perspective on inflammatory events after pe- demyelination. Brain 134:3590–3601, 2011 ripheral nerve injury. J Neuroinflammation 8:110, 2011 34. Wargo CJ, Gore JC: Localized high-resolution DTI of the 15. Grant GA, Britz GW, Goodkin R, Jarvik JG, Maravilla K, human midbrain using single-shot EPI, parallel imaging, Kliot M: The utility of magnetic resonance imaging in evalu- and outer-volume suppression at 7T. Magn Reson Imaging ating peripheral nerve disorders. Muscle Nerve 25:314–331, 31:810–819, 2013 2002 35. Zhou Y, Kumaravel M, Patel VS, Sheikh KA, Narayana PA: 16. Grant GA, Goodkin R, Kliot M: Evaluation and surgical Diffusion tensor imaging of forearm nerves in humans. J management of peripheral nerve problems. Neurosurgery Magn Reson Imaging 36:920–927, 2012 44:825–840, 1999 17. Howe FA, Filler AG, Bell BA, Griffiths JR: Magnetic reso- nance neurography. Magn Reson Med 28:328–338, 1992 Disclosure 18. Krause TL, Bittner GD: Rapid morphological fusion of sev- The authors report no conflict of interest concerning the materi- ered myelinated axons by polyethylene glycol. Proc Natl als or methods used in this study or the findings specified in this Acad Sci U S A 87:1471–1475, 1990 paper. 19. Kretschmer T, Heinen CW, Antoniadis G, Richter HP, König RW: Iatrogenic nerve injuries. Neurosurg Clin N Am Author Contributions 20:73–90, vii, 2009 Conception and design: Riley, Boyer, Kelm, Sexton, Does, 20. Lee JY, Giusti G, Wang H, Friedrich PF, Bishop AT, Shin Thayer. Acquisition of data: Riley, Boyer, Kelm, Sexton, Pol- AY: Functional evaluation in the rat sciatic nerve defect lins. Analysis and interpretation of data: Riley, Boyer, Sexton, model: a comparison of the sciatic functional index, ankle Pollins, Dortch, Does, Thayer. Drafting the article: Boyer, Kelm, angles, and isometric tetanic force. Plast Reconstr Surg Thayer. Critically revising the article: Riley, Boyer, Kelm, Pol- 132:1173–1180, 2013 lins, Dortch, Does, Thayer. Reviewed submitted version of manu- 21. Leemans A, Jeurissen B, Sijbers J, Jones DK: ExploreDTI: a script: all authors. Approved the final version of the manuscript graphical toolbox for processing, analyzing, and visualizing on behalf of all authors: Riley. Statistical analysis: Riley, Boyer, diffusion MR data. Proc Intl Soc Mag Reson Med 17:3537, Kelm, Does. Administrative/technical/material support: Shack, 2009 (Abstract) Thayer. Study supervision: Boyer, Shack, Thayer. 22. Lehmann HC, Zhang J, Mori S, Sheikh KA: Diffusion tensor imaging to assess axonal regeneration in peripheral nerves. Supplemental Information Exp Neurol 223:238–244, 2010 23. Li X, Chen J, Hong G, Sun C, Wu X, Peng MJ, et al: In vivo Previous Presentation DTI longitudinal measurements of acute sciatic nerve trac- Portions of this work were presented in abstract form at the tion injury and the association with pathological and func- American Society for Peripheral Nerve (ASPN) Annual Confer- tional changes. Eur J Radiol 82:e707–e714, 2013 ence in January 2014. 24. Liang D, Bhatta S, Gerzanich V, Simard JM: Cytotoxic edema: mechanisms of pathological cell swelling. Neurosurg Correspondence Focus 22(5):E2, 2007 D. Colton Riley, Department of Plastic Surgery, Vanderbilt 25. Morisaki S, Kawai Y, Umeda M, Nishi M, Oda R, Fujiwara University School of Medicine, 1161 21st Ave. S, MCN S-221, H, et al: In vivo assessment of peripheral nerve regenera- Nashville, TN 37232. email: [email protected].

Neurosurg Focus Volume 39 • September 2015 9

Unauthenticated | Downloaded 10/10/21 08:48 PM UTC